What if your ai coding agents could talk to each other? We put Claude Code and Gemini CLI in a shared chat and let them plan, build, and review a full app autonomously, this is what vibe coding looks like with multi-agent coordination.

🔗 Links
- Repo - https://github.com/bcurts/agentchattr
- Our Repo With Added UI - https://github.com/ailabs-393/agentchattr
Community with All Resources 📦: http://ailabspro.io/
Video Code: V50

In this video, we explore Agentchattr, an open-source tool that lets you run multiple ai coding agents side by side and have them communicate in real time through a shared chat interface. Instead of switching between tools yourself, you assign roles to each agent and let them coordinate autonomously across different channels.

We walk through the full setup, from initializing agent sessions with tmux to configuring permissions and MCP tools so agents can operate without constant manual approval. We cover how to structure your project with planning templates like a PRD, backend spec, and UI spec, and how to use an agents.md file so that both Claude and Gemini follow the same guiding instructions regardless of their native config files.

For anyone looking for the best ai for coding workflows that involve multiple models, this is one of the most interesting approaches we've tested. The tool supports coding with ai across Claude Code, Gemini CLI, Codex, and open-source models like Kimi and Qwen, making it one of the more flexible ai coding tools available right now. You can use an expensive model for planning and a cheaper one for implementation, which is a practical way to manage costs while still getting the best coding ai output where it matters.

We demonstrate a full planning-to-build pipeline: one agent creates the PRD, another handles the UI spec, and a builder agent implements the backend, all coordinating through the chat. The tool also includes structured review modes where agents challenge each other's work across three phases, catching gaps that a single ai for coding session would miss. The planner then acts as an orchestrator, delegating tasks and keeping agents from overwriting each other's files.

If you're into ai vibe coding or agentic ai workflows, this is a practical look at what multi-agent collaboration actually looks like today, the wins, the edge cases, and the setup needed to make it work. We also discuss how using dedicated work trees with one agent handling the merge and review would make ai automation with multiple agents much smoother for complex projects. Whether you're using claude code or opencode or gemini cli, coordinating them in a shared session opens up new ways to build.

Hashtags
#claudecode #claude #vibecoding #aiautomation #coding #opencode #agenticai #geminicli #aicoding

AI LABS

Agent Chatter is an open-source chat interface that enables real-time coordination between multiple AI coding agents (Claude Code, Gemini CLI, Codex, and open-source models). It uses a shared multi-channel chat to let agents communicate autonomously, assign roles, and divide work. The post walks through setup using tmux, configuring permissions and MCP tools, creating planning templates, assigning agent roles, and using a multi-phase planner/challenger/synthesizer review mode. Key workflow tips include pre-initializing your framework, using dedicated git worktrees to avoid agents overwriting each other, tuning the loop guard, and letting a planner agent orchestrate parallel frontend/backend implementation.

This is the Ultimate AI Agent System